amingclawdev/aming-claw

Aming Claw

Push, not pull: a guided runtime for AI agents.

Aming Claw is a local-first governance layer for AI-coded development. It keeps an agent on track by pushing the current role, contract, next legal action, required evidence, and verification gate before the agent acts.

That is the core move from the public article: Push, not pull: your AI agent shouldn't fetch its own context. Search, grep, RAG, and repo maps still matter, but they are fact-layer tools. They should not be the control loop for an agent that is already drifting.

Status

Aming Claw is a V1 MVP and active dogfood system. The guided-runtime direction is mid-migration: more work is moving from agent-pulled context into runtime-pushed contracts, gates, and evidence. The current evidence is observational, not a validated benchmark claim.

Today it is useful for local AI development governance:

• Local dashboard for backlog, activity, graph, review queues, and runtime evidence.
• MCP tools for agents to query live project state instead of relying on chat memory.
• Commit-bound project graph and event-sourced runtime/timeline evidence.
• Onboard service that selects the role, work type, token refs, and next contract path.
• Direct-fix and parallel-work paths for governed implementation.

Install

Use the host you work in, then reload or open a new host session after install. Plugin files on disk do not automatically appear inside an already-running AI host conversation.

Codex One-Shot

Ask Codex:

One-shot install and open dashboard for Aming Claw from https://github.com/amingclawdev/aming-claw

Install-only refresh:

aming-claw plugin install https://github.com/amingclawdev/aming-claw
aming-claw plugin doctor

Claude Code One-Shot

Paste this into Claude Code:

Ins